JP3681754B2 - Data management system configuration method - Google Patents

Description

The present invention relates to a method for configuring a data management system for accessing data in a distributed data structure.
Many different configurations provide complex distributed data structures. For example, complex distributed data structures are provided by different nodes in a parallel processing computer. On the other hand, in a communication network, a complex distributed data structure is provided at different switching points in the network, for example, in a local exchange of a public network. In certain embodiments, the present invention can be used to provide routing information for routing calls to users who can change their location with respect to their logical or geographical location in the network.
The services available through communication networks are becoming increasingly complex. It has become important to introduce management systems for related service mechanisms that can be selectively used by customers but do not introduce excessive computation or communication overhead. This example is known as a personal telephone numbering method. In these methods, a telephone number is assigned to each user or group of users that does not change when the user changes his position in the network.
Clearly, for services such as personal numbering, it is important to be able to easily change location data. From a network perspective, location data is volatile, but its current version needs to be always accessible quickly.
Another aspect of network services such as personal numbering is that it is often important to be scalable. That is, it needs to be able to scale to a larger number of users, perhaps an infinite number of users. For example, it is often important to be able to grow a network to support about 1 million users. Also, these users tend to generate variable loads. For example, a user who arrives at the office wants to move his / her personal number from his home position to his / her office position. This generates millions of transactions in a few hours. Assuming that the fixed network supports a complete mobile service that is equivalent to that of a cellular radio network, the overhead conveyed by the fixed network in updating location information results in millions of transactions per hour.
One option is to maintain a central database that holds data related to services such as personal numbering. Traffic that must be routed to a destination in the network accesses this database to obtain a current location for the destination. However, this type of system creates significant overhead and the network must carry traffic that simply accesses the central database to update location data or download the current version. It is clear that the configuration based on central control is vulnerable to overhead.
Another method used in cellular (mobile) telephone networks is to provide each user with a home register known as a “home location register”. Each home register holds data related to the user, a type well known as a user profile in intelligent network technology. That is, the home register indicates which service the user is subscribed to, and provides detailed information such as the day time information for a “call diversion” type service that can be applied differently at different times of the day To do. The home register also has position data for the user. The plurality of home registers together provide an effective distributed data structure and is arranged such that there is a mapping between the user's normal location and the location of the user's home register in the network. However, since all calls go automatically to the called user's home register, considerable overhead remains when dealing with users away from their normal location. Furthermore, changing the user position means that the position data in the home register must be updated, which increases the signal load. In addition, there are problems when the size of the customer base increases. The design response to an increasing customer base is to reduce the local cell size. However, not only does the signal load increase dramatically according to the square law, but also the complexity of the management procedure increases. Providing data integrity and strength is an important issue.
A slightly more complex “Home Location Register” (HLR) configuration has been proposed, which is provided with additional registers, known as “visit location registers”. This repeats the characteristics of the HLR but is placed elsewhere in the network to handle call requests when the called user is away from his normal location. Depending on the location in the network where the incoming call occurred, the incoming call is related to any register in these environments. However, if the call request is routed to the HLR, the call request is simply sent to the visit location register.
According to the present invention,
Comprising a hierarchy of nodes having communication links between the nodes, the hierarchy extending from a root node to a plurality of end nodes, the plurality of end nodes providing a distributed data structure and being stored at the end node through the hierarchy A route to a particular data element is identifiable by pointers stored at nodes along each route, and a request to access the particular data element has a pointer associated with the particular data element Trigger a search message sent from the end node to the root node through the hierarchy until the node is reached, after which the search message includes the data element along the associated route to the end node Data that is sent to access data elements stored in a distributed data structure A method for recovering a data access system in a method for configuring the access system,
(A) detecting a faulty part of the hierarchy that affects communication between at least one child node and each parent node of the hierarchy;
(B) establishing a first other communication link between the child node and another node, the another node being each second parent node for each child node;
(C) establishing a second further communication link between the second parent node and the first parent node;
(D) instructing the first parent node to send a message directed to the child node to the second parent node that sends the message to the child node;
(E) recovering a data access system comprising the step of the second parent node periodically updating other nodes in the hierarchy with respect to a new location when the child node is in the hierarchy A method is provided.
A second separate communication link is established, and then the hierarchy is periodically updated with the new position of the child node in the hierarchy, thereby distributing the signals necessary to update the hierarchy in time and Can be limited to periods when the “load” is low. Another communication link satisfies a request for a child node from a node that has not been updated for the new location of the child.
Individual numbering has already been mentioned above. This is expected to be an important component in future communications services. Such systems assign numbers or addresses to individual users (or possibly to groups of users). Traffic directed to that user is then routed to that person no matter where the person is located in the network. There is always a current hardware address associated with the user so that the call is routed to the destination, but the hardware address is simply a translation of the assigned personal number, and when the associated user moves around the network Be changed.
In the embodiment of the present invention, the hardware address is represented by a data element included in the end node. The personal numbering system wants incoming calls to you to be redirected to hardware addresses in the network handled by different local exchanges. The user updates the end node in the new local exchange so that the associated hardware address is associated with the user's personal number. The end node then updates the associated pointer at the node in the hierarchy to inform the route to the new local exchange of the call for the user's personal number and the user's previous end node where the user was located. It is preferable to instruct to cancel the hardware address for the personal number. This means that the location data for the user is moved and changed. However, a data access system according to an embodiment of the present invention tracks user movement and provides an updated version of the hardware address. This tracking is triggered at the local exchange.
In accordance with an embodiment of the present invention, a data access system can automatically reconfigure itself and / or reconfigure its information “space” upon demand or upon failure. Can do. For example, if an embodiment of the present invention has a hierarchical nature, the "loss of data" can be potentially found and reconstructed if necessary in the reconfigured system. The “floodfill” search procedure can be rapidly extended across all data access systems.
Embodiments of the present invention compare against service routing issues such as personal numbering where routing must be changed at the level of detail in the network in real time, even in a fixed public switched telephone network (PSTN). Provide a simple solution. Otherwise, in PSTN, development and change pose significant difficulties for routing personal numbering.
A network management system for personal numbering will now be described as an example of the present invention by way of example only with reference to the accompanying drawings.
FIG. 1 shows a simple four-level binary tree configuration of a network management system.
FIG. 2 is a flowchart showing a simple information retrieval procedure executed by the node in the configuration of FIG.
FIG. 3 shows bridging of two network management systems according to the configuration of FIG.
FIG. 4 illustrates an extended network management system according to an embodiment of the present invention.
FIG. 5 is an example of a response to a failure in the network management system according to the configuration shown in FIG.
FIG. 6 shows an information layer including a plurality of network management systems according to the configuration shown in FIG.
FIG. 7 is a diagram used in the description of the recovery mechanism according to the present invention.
FIG. 8 is a data flow diagram.
FIG. 9 is a block diagram of an intelligent network architecture that can implement the network management system of FIG.
In the personal numbering system in the communication network, the characteristics of the service requested by the user are:
1. The ability to program a local terminal device such as a telephone to be a destination device for personal phone numbers,
2. Ability to return to a predetermined location in the network, such as a home phone, as a destination device
3. The ability to select a destination device for a personal telephone number at will in the network.
This means that the network and associated management system must be able to work together to route incoming calls to the “moving target”. The basis of the design of the network management system according to the embodiment of the present invention is to logically separate the information management task from that of the network base, which can easily be extended while handling the control of the distributed database. Realized in signaling network. Typically, in personal numbering, a system that includes a signaling network with an analytics database will have positioning information, which is usually a hardware destination address in personal numbering, but once it is on the network infrastructure or bearer network Bring control back to effect.
The advantage of this separation between the bearer network and the management network is to facilitate the coupling of end-user control systems with networks such as mobile communications, internal business, computer networks, etc. Since the management network is generally an independent conceptual structure, it can be physically handled as an independent structure and can be directly provided on the current system. Thus, the management network does not replace the current system, but rather can augment it functionally. In addition, the management network can span several heterogeneous networks, which can be unified.
Referring to FIG. 1, the simple basis on which the network management system operates is nodes 1, 2, and 3. Nodes 1, 2, and 3 are connected in a hierarchical manner to form a tree-like signal network (a tree-like network is a logical structure that can be realized in an actual physical network, Keep in mind that a good network doesn't necessarily have to be a tree). Every node 1, 2, 3 has at least one parent, with the exception of one node 2, which forms the "root" of the tree. The root node 2 is special only in the sense that it does not have a parent. This structure, consisting of nodes with only one parent, has the advantage of simplicity, and in the case of multiple parents, more complex control logic is used to handle the extra decisions to be made. I need. However, structures with multiple parents are more robust in physical and logical failures. Embodiments of the present invention are not limited to use with a single parent structure.
In general, the nodes 1, 2 and 3 have one or more children except for the end node 3 which directly handles the bearer network, and these are usually located in the local exchange. In order to simplify the following description, it is assumed that these end nodes 3 need not have any children and have no children.
Referring to FIG. 1, the general principle of a routing network comprising nodes 1, 2, 3 and links is that each end node 3 can potentially contain a current hardware address for one or more users. (Or something that can be provided). When an incoming call request is received by the network to set up a call for a destination user, the routing network is used to find the current hardware address associated with that destination user. The routing network can do this by finding the location of the destination user's current hardware address via any one of the end nodes 3, but the node above the correct end node 3 in the hierarchy. This is because the correct end node 3 is informed in the routing network by the pointer of the correct end node 3 in 1 and 2. Regardless of which end node 3 the call request entered the network, the call request is simply sent through the hierarchy until the call request encounters a pointer clue to the correct end node 3. That is, a call request is sent over the network until a call request is met with nodes 1 and 2 that contain pointers towards the correct end node 3. The pointer directs the call request to the next node in the clue, and similarly back down the hierarchy towards the end node 3 that can provide the current hardware address of the destination user.
All nodes 1, 2, 3 can have the same data structure and processing unit. Data is present in the data structure only if that particular node is responsible for directing the call request to the destination user. Its role is
i) simply send a call request to its parent node;
ii) if the node is not at the destination but is in the middle of the path to the destination, provide a pointer to the destination and send a call request to the node indicated by the pointer;
Or
iii) If the node being targeted is the end node 3 for the destination user, then providing the current hardware address of the destination user.
The information present in the data structure of the node is a pointer, which contains a “key” and a link identifier, and in the case of end node 3, it contains the current hardware address (other information present is This is a list of parent nodes to use in case of failure, except for the root node, which will be discussed further below (alternatively, a node could look for a new parent). So far only the end node 3 has included the current hardware address, but generally all nodes will be able to include it. As will be understood later in this description, this is because end node 3 becomes intermediate node 2 and vice versa when the routing node is expanded or contracted. The “key” generally identifies the user associated with the pointer. For example, the key is the user's personal number. The link identifier identifies which down link to which the call request should go to the next layer of the routing network.
A routing network as previously described clearly does not provide a “home location register” for any user and therefore avoids the additional signal load associated with call requests to or through the home register. It is noted that it is a thing.
Using the signaling network shown in FIG. 1, a user such as “P” is typically handled by an end node 3 such as the nearest local exchange. This means that the local network's current network hardware address is stored in the local exchange. When user P decides to move from being handled by one node A to being handled by a different node B, he causes a change. User P makes changes by accessing Node B, and Node B modifies its data to indicate that User P can be obtained directly from Node B without additional routing. Node B then sends a change command to its parent node D, so that parent node D changes its data to route to node B instead of routing calls destined for user P to node A. . Node D acting as a parent to node A informs node A that node D is no longer routing to user P and therefore any calls for user P should be sent to parent node D. .
At this stage, it can be considered that the user P has substantially become the user P ′. The data structure on the network is not substantially changed, and only the nodes A and B together with the parent node D are affected.
Only node B may be able to find user P. This is at a stage where node B modifies its data to match user P, but parent node D is not making the resulting change. However, this is a very short time.
If user P moves to a farther node, for example to node C in FIG. 1, the change will necessarily extend farther. Initially, node C changes its database to match user P, and informs its parent, node H in FIG. 1, that the change has occurred. The change propagates through the tree, such as node C teaching node H, node H teaching node G, node G teaching node F, and tree roots. Node F then “destroys” the previously existing route to user P by informing his child node E that he no longer routes calls destined for user P to his child. In turn, node E communicates to its child node D, which communicates to its child node A.
In the worse case that can occur, these changes are even more, but the number of changes involved is 2 log, where N is the total number of nodes in the tree. ₂ (N) only. What is important is that the change is local and triggered by communication between adjacent nodes. This makes the updating of the distributed routing table very efficient and can be done with current technology.
Referring to FIG. 2, if a call request is made to establish a connection to a destination user in the bearer network, the routing network is triggered to locate the destination user. The routing process is the same for each node that receives a call request, and the node makes one or more series of decisions and functions based on the results of these decisions.
The routing process at any one node that receives the call request is seen in the flowchart 20 of FIG. The node's first response, step 22, is to examine its data structure for information about the destination user. The three possible scenarios are as follows:
1. The target node is not at the destination user, nor is it located in the middle of the path to it.
In this case, in step 22, the node knows that no information exists. Since it is sent towards the root of the call request network, in step 23 the node then checks whether it is a root node. The normal situation is that the node is not a root node as in step 24, and a call request is simply sent to the parent node. As a result, the communication request creates an upward path toward the root node 2 via the network of nodes.
2. The targeted node is in the middle of the path to the node where the destination user is, but is not itself the local exchange of the destination user.
In this case, the node knows in step 22 that it holds information about the destination user. Therefore, the node starts checking whether it is the node where the destination user is located. That is, as in step 26, the node checks whether it is a local exchange. In this scenario, the answer is negative as in step 26, and the node sends a communication request to the child node indicated by the existing information. Therefore, the communication request at this stage is sent to the network root node and again down towards the destination user (however, it should be noted that it is not always necessary to route the communication request through the root node. is there).
3. The node is a node adjacent to the destination user.
In this case, from steps 22 and 25, the node knows that it has information about the destination user and is also a local exchange. This node therefore downloads information to the bearer network that can set up the connection. In other words, this node downloads the location of the caller who generated the communication request and its own location information. As in steps 27 and 28, the routing network can leave the operation and the bearer network routes the call appropriately on its own.
There is one other result that has not been discussed previously. That is the case when the node establishes that it has no information about the destination user but is the root node 2 and therefore has no place to direct the communication request. As a result, information is lost, indicating that instead of sending a communication request, node 2 causes a data search as in step 29.
A system based on a tree network as previously described uses a parallel concept. Each node 1, 2, 3 manages a database that is updated in response to changes caused by its neighbors. Hierarchical construction imposes limitations on the need to change information. It is possible to propagate the requested data changes throughout the system, but the structure is designed so that the changes have limited results. This level of parallelism is evident from the geographic decomposition, but to be precise, the same decomposition can be applied within nodes 1, 2, and 3.
An important feature is that changes in the distributed data structure are handled as locally as possible. When the piece of information is updated, most of the nodes 1, 2, 3 are not involved. This has the effect of distributing the load for updates across the network. The result is that the signaling of information continues to be distributed to the network even when the system becomes larger. In the case of a binary network, as described with reference to FIG. 1, the signal load per node 1, 2, 3 is independent of the number of nodes 1, 2, 3 in the network.
Each of 1, 2, and 3 may be provided by a distributed memory parallel computer. Each processor in the computer can be directly assigned to one of the communication nodes 1, 2, and 3 described in the above example. The computer control system can manage the database in the same way that the nodes 1, 2, 3 operate. This is the node representing the local switch level, and each “leaf” of the tree, which is the lowest and most frequent node 3, solves the problem in the same way as the higher nodes in the network and is therefore the same This means using a control processor.
This is particularly advantageous in that each node can potentially be used to control the lower nodes of the network, always using the same control process. Thus, as previously described, to extend the scope of a routing network, a new network is created based on the same architectural principles, a new root node is created, and the root node and network of the original network are “ You can “degrade” below the new root node.
Referring to FIG. 3, the original network has a root node “R ′”, and the new network has a root node “R ″”. A new root node “R” is then entered as a new layer, replacing the nodes that were previously root nodes R ′ and R ″, and the new root R functions as the root corresponding to the two networks. Each of the old root nodes R ′ and R ″ must be updated to record the fact that they are no longer root nodes. The system will then function properly. The same principles can be used to extend current routing networks that are geographically heterogeneously distributed to meet growth and demand. Thus, node 3, which previously represented the local exchange, is converted to a higher node 2 in the routing network by simply adding another network below it. This is illustrated in FIG. 4, where the node 40 that was previously at the lower most level of the routing network has been converted to the role of an intermediate node above the six additional nodes 41, 42 to represent the local switch. The lower most nodes that are now are the bottom two layers.
In each case, as long as each node whose status has changed is informed, the routing network is effectively in a calm state after the configuration change.
Referring to FIG. 5, the routing network can be designed to withstand signaling links or failure of nodes 1, 2, 3 or most of the network.
The method of detecting link failure is either because the connecting node failed to use the link or because the network management mechanism found the failure. An example of the latter is the “heartbeat” signal supplied in a conventional C7 signaling system.
For example, in the simplest situation where there is a failure on one link between end node 3 ′ and intermediate node 2 ″, end node 3 ′ normally sends to its parent node 2 ″ on the other side of the failed link. Can no longer respond to incoming communication requests. In a normal routing network structure, there is no alternative link from its end node 3 ′ to the original parent node 2 ″ or some other parent node.
In order to reconfigure the routing network under a link failure, each child node 2, 3 has a list of parent nodes, which causes each child node 2, 3 to attempt to establish reconnection in the routing network. be able to. Therefore, the child node 3 'affected by the failed link issues reconnection requests to each potential parent node in the list in turn until it finds the location of the functioning parent node with spare capacity. To do. Thereafter, the reconnection of the route to the child node 3 ′ via the network is set up in the same way that the reconnection is established when the user moves from one end node 3 to another.
Since a reconnection request must be issued for all pointers contained by an associated child node, in practice this reconnection execution depends on how high the associated node is in the hierarchy. It becomes complicated.
In FIG. 5, a new link 50 has been established. However, the routing network has a logically defined structure in which links are identified by addresses at each node, for example, so that a new link 50 creates a new link 50 even if it straddles two layers of the hierarchy. There is no difficulty.
On the other hand, this is the case when a communication request destined for the end node 3 'enters somewhere else in the routing network and therefore needs to be routed to or through the child node 3' below the failed link. It may be. If the child node 3 'has already triggered the execution of the reconnection using the new parent, then the network functions in the normal way. However, performing reconnection causes a delay depending on the number of routes that need to be reestablished. If this execution is not complete, the incoming call request may reach the intermediate node 2 ″, the parent to the end node 3 ′, but not the end node 3 ′. As efficiently as a child node can wake up a new link 50 to reach a new parent, the intermediate node 2 ″ cannot wake up a new link 51 to reach the child node. This is because the new child node is only concerned with the selection of the pointer that is conveyed in the intermediate node 2 ''. In this case, it is more efficient for the intermediate node 2 ″ to issue a “floodfill” search request to locate the new parent 2 ′ of the child node.
A floodfill search request is simply issued from one node to all other nodes to which it is connected. When a flood fill search request is received, each node checks to see if it contains related data or related pointers. If so, the node responds to the request. If not, the node sends a floodfill search request to all connected nodes except the node for which it received this request (a floodfill search request actually sends a request, for example) Can be limited by limiting the number of times it can be).
The intermediate node 2 ″ can therefore locate the new parent node 2 ′ for the end node 3 ′ and re-establish all relevant routes.
The mechanism described above thus creates a new link 50 in the network that effectively “bypasses” the failure.
The same mechanism can of course be used, whether it is a link or a node that has failed.
There is a second mechanism for re-establishing the network after a failure. A flood fill search request for specific information can be issued simply to the entire network. By applying this method, a large-scale network reconfiguration can be achieved.
The use of flood fill search request techniques creates significant signal overhead and is not desirable for less robust networks, i.e., those that are prone to failure.
When a node establishes a link to a new parent, the node needs to send a “new” location to the new parent node. This should be done as soon as possible if the request is sent to the node without delay. The need to update the network after an address change followed by the establishment of a new connection between the node and the new parent causes a great deal of signal movement. This creates a problem with the network's ability to route requests between nodes, especially when there are multiple reconnections during a given time interval, as in any realistic network. .
The present invention has been made to solve this problem.
FIG. 7 illustrates the recovery method. In the routing network 70, there is a link failure in the link 71 between the nodes 72 and 73.
The first step in the recovery mechanism is a detection step 80, as shown in FIG. This step is executed by the node 73 and detects that the link 71 between the node 73 and its parent node 72 is broken. Node 72 performs this step as well.
The second step 81 is performed by the node 73, which is the establishment of a new connection 74 to the new parent 75. A new parent is selected from the list of potential parents held by node 73.
The new parent node 75 then establishes a direct link 76 to the old parent 72 in the step represented by box 82.
The old parent 72 is then requested by the new parent 75 to update its address for the node 73, and all requests received by the old parent 72 are sent to the new parent 75 via the direct link 76 and to the new connection 74. To the node 73.
Therefore, the link failure is resolved at this stage, and the node 73 becomes a part of the network 70 again.
The next step, as represented by box 84, is for the new parent node 75, which is to periodically update other nodes in the network for the new location of node 73. Instead, this process is performed “on demand”. That is, when a request for a node 73 is received from a node, the new parent 75 sends an update location message to the sending node. This can depend on whether the request is a repositioning to a call connection request. Alternatively, this can be performed according to a predetermined pattern or according to network activity. That is, the update process can be performed during periods when there is not much signal movement in the network.
The previously described network management system or data access system provides a relatively reliable data structure because the data structure is distributed. The information accessed, in this case the hardware address for the network user, is not kept in one particular location. In the first example, there is only one route through the system to the data element of that information, but faults in the route can be easily handled as explained earlier. An additional feature provided is to increase the security of the connection to the user, a backup data storage point, a “ghost” end node that holds the current hardware address for the user. Since the location can be located by a flood fill search request, reconnection for the user can still be established.
When a flood fill search request is issued, a relatively high level of signal traffic occurs. To minimize usage, flood fill search requests may only be used in the case of data tampering or requests. In a tree-like hierarchy as previously described, the path to the local exchange via the routing network, ie to the end node 3 in direct contact with the target user, is always 2 log. ₂ (N) ^-1 Make this search procedure relatively efficient, even less.
In addition to searching, floodfill search requests remove any reference to specific data on the nodes through which they pass, thereby eliminating any bogus routes or data paths in the network. The only exception to this is the case at the local exchange or end node 3 where the information is found. In this case, the location recording process can be initiated so that the data (or the route to the data) can be reconstructed. In addition, a call configuration request can be sent to the bearer network to ensure that the initial request is handled. In the worst case, this entire process for connecting calls is 4 log. ₂ (N) ^-1 It takes a step and is exactly twice that of a complete network.
Large scale damage can be reconstructed in this way, but it is more efficient to use a start routing request to reconstruct directly. The combination of both processes leads to a fast recovery process where the necessary data is created immediately and unrequired data can be slowly recovered when the network capacity allows.
A management system of the type described can be easily extended. Even in the simplified case represented, ie in the tree structure, the generalization to the mechanism for retrieving distributed information is clearly the next step. Even in an environment that dynamically repositions information and introduces multiple copies, a hierarchical control system provides the earliest access to the closest source of information required. This is true even if the first node does not know the data for a point that does not recognize the database key or location address.
This can best be understood from a distributed database perspective by the operation of the personal telephone numbering system described above. The fact that the user's location is continuously monitored and updated makes the internal structure of the management network information an internal structure of an automated database. Adding the fact that there is physical distribution of data sources, ie user movement, that physically distributes data, this minimizes the number of transactions that each node must handle. The end result is a distributed data structure that can be accessed from any of its remote terminals as if it were a close database, while data transactions are updated in parallel.
In direct analogy, information such as that stored in a library is often placed at many geographic locations. In order to access such information, the location where the information is stored must be known. This is a trivial problem in a static environment. However, problems can arise. When there is only one copy of each piece of information, multiple simultaneous accesses to a given piece of data become a problem. The obvious solution to this is to move the data and make multiple copies of this data. That is, popular data is placed at several geographically separated locations. This reduces access contention and network load. Again, once this is done, the data source is always at the same location, so there is no need to worry about finding the location of the data source from a predetermined location. However, continuous changes are likely to be required in the real world.
It is well known that the popularity of information fragments is transient. This is directly related to the fact that the information is volatile. Fragments of information are often valuable only when they are new, and often are worthless as time starts to pass. Due to the change in the relative value, the position of the data viewed from the viewpoint of some predetermined terminal changes with time. Informing all users of the system about these changes in location puts a heavy cost on the network load. Therefore, there is a need for a system that can change the location of information without notifying the user and at the same time maintain the user's access to the information. This is exactly what the hierarchical control structure of the embodiment of the present invention can do, and as a result it is quite versatile.
A “key” to a database can literally be thought of as a key. This is simply a structure by which information is referenced. Similarly, references at the end of the paper can be seen as a key. These do not necessarily need to be understood, but can be incorporated elsewhere to be read or at least partially read. In the example of a paper with a reference, the librarian is consulted and the librarian teaches the person how to locate the paper in the library or tells the person how to go to the correct type of library.
There is a unique hierarchy even at this level of key type. Different control layers can be overlaid on a common distributed data source.
Referring to FIG. 6, there is a common distributed data source at end node level 60 in three different control hierarchies. The three control hierarchies 61, 62, 63 are structurally similar, each having a root node at the “top” and accessing an array of nodes in a common distributed data source 60. Different control layers can be easily connected to the extent that they can compare and share management information. In FIG. 6, some of the child nodes, in this case end nodes, contain only one type of data storage, while others contain three types of data storage, It does not mean that these “single children” cannot access data from all three types.
There is no reason why this scenario cannot be developed into a general database structure. For example, it is entirely possible to use such a system to integrate different databases. Physical distribution presents the same data management difficulties as algorithmic distribution, that is, when the algorithmic control of data is separated into different structures. This is usually a reference to two different database programs such as Sybase and Oracle. Such programs already share a common interface language and respond to the same form of data requests from remote terminals. The data request is first structured into this common language by the application program, and the received data is converted into the relevant format. Each database program can be viewed as a physically separate, ie physically separate, data storage device. As shown in FIG. 6, all that is needed to unify the two systems into a hierarchical control structure is a simple conversion of requests from the hierarchical control structure into related database requests. Interface program. This makes it possible to apply a hierarchical control structure to the current database system in a very simple way.
A hierarchical routing structure according to an embodiment of the present invention provides a general service that allows data retrieval from some part of a hierarchical routing network without necessarily knowing the location of the data source. Can do. This functionality can simply be viewed from the user's perspective, for example, as a service structure that can be accessed at any point. This can be treated as a system that allows the user to issue a request from any physical point, and the system resolves the request to the user.
The user terminal has a direct connection to the hierarchical control network. User terminals can issue requests to the system, and these requests take the form of queries about data contained in the data space of the routing network. In a particular embodiment of personal numbering, this information is the user's current hardware address identified by the dialed telephone number. In practice, the routing network typically does not return the current hardware address, but instead proceeds to trigger an automatic connection of the call from the location of the found current hardware address. This can be applied in other situations. That is, any application in the routing network has some predetermined response when a piece of information is found. The response is to return the information to the person who inquired it, but the return of this information can be simplified by considering it as a separate task, and the actual transfer of information between two related addresses The conventional transfer process may be used. The task of the routing network is simply to locate the information. Different mechanisms can be used to handle information once discovered.
As can be seen, this movement is independent of the position of the user terminal. Thus, the routing network can be viewed as a database of non-local information. More importantly, this database does not need to hold data in a particular location. If another location is found to be the more optimal storage location, the information can be moved without affecting any of the user terminals or the data access key used by the user to retrieve the information. it can.
Referring to FIG. 9, considering the application of routing network to personal numbering according to the embodiment of the present invention described above, the routing network is generally provided at the service control point 90 of the intelligent network architecture. Can do. In an intelligent network architecture, the connection is usually established by a service switch point 91 using routing information obtained from the associated service control point 90. Instead, if the intelligent network architecture is quite flexible, the routing network replaces the function of the service control point 90 and is located elsewhere in the intelligent network architecture. Hardware that provides the functionality of a routing network is known and already used in intelligent network architectures. For example, the associated hardware is provided by a UNIX workstation such as Sun Spark 5 with a gigabyte capacity hard disk.
Important features that can be utilized by embodiments of the present invention can be listed as follows.
1. Optimization for geographic demand, for example by adding layers to add depth to dense areas in a personal numbering service network.
2. In the problem of libraries, an efficient arrangement of demands can be provided, for example, in services that are literally libraries, or in services such as video on demand that are obviously of the library type.
3. Even if the signal load is kept low and the network as a whole is expanded, the signal load can be maintained so that the signal load at any point is constant. This aspect is because the network can generate additional local layers as demand increases.
4). Embodiments of the present invention can rely on conventional technology and software already in use, and therefore do not require the development of a separate platform.
5. strength. Although not previously described, the strength can be improved by increasing the number of parent nodes for one child node. That is, there may be two or more parent nodes associated with one child node. This is illustrated in FIG. Another aspect of strength is the natural recovery of the routing network.
6). In accordance with an embodiment of the present invention, a new layer can be introduced at the top of the routing network, so the common root node superimposed to transform the original root node of the different network into a child node allows The routing network can be linked. This means that networks can be combined and expanded to different national regions.
7. This type of routing network allows for the introduction of new services, since nodes do not need to be updated for new services, and nodes only need to provide routing and know the data for which routing information is generated. It does not have to be. In conventional networks, nodes contain number conversion files that must be updated when new information is introduced into the network. In accordance with the present invention, this requirement is reduced or eliminated in network operation.
The hierarchical structure described above is typically a binary tree structure, but they need not be binary or tree-like. The structure can for example be considered as three-dimensional rather than two-dimensional and / or may comprise multiple links instead of the binary method. However, an important aspect is a hierarchy where there are fewer “upper” layer nodes and more information than “lower” layer nodes.
The layout of the structure used can be optimized to suit the task that needs to be performed.
Although the embodiment relating to the personal numbering service described above has been described in connection with a fixed network, it can also be applied to a network that is an integration of a fixed network and a mobile cellular network. Indeed, the embodiment is appropriate for a cellular network alone.

Claims (10)

Comprising a hierarchy of nodes having communication links between the nodes, the hierarchy extending from a root node to a plurality of end nodes, the plurality of end nodes providing a distributed data structure and being stored at the end node through the hierarchy A route to a particular data element is identifiable by pointers stored at nodes along each route, and a request to access the particular data element has a pointer associated with the particular data element Trigger a search message sent from the end node to the root node through the hierarchy until the node is reached, after which the search message includes the data element along the associated route to the end node Stored in a distributed data structure by means of the data access system being sent A method of accessing that data element,
The method of accessing the data element includes a recovery method;
The recovery method is:
(A) detecting a faulty part of the hierarchy that affects communication between at least one child node and each parent node of the hierarchy;
(B) establishing a first other communication link between the child node and another node, wherein the other node is a second parent node for each child node;
(C) establishing a second further communication link between the second parent node and the first parent node;
(D) instructing the first parent node to send a message directed to the child node to the second parent node that sends the message to the child node;
(E) accessing a data element including the step of the second parent node periodically updating other nodes of the hierarchy with respect to a new position of the child node in the hierarchy. The method of claim 1, wherein each of the data elements includes an address identifying a data storage location in a data structure. The method of claim 1, wherein each of the data elements includes an address that identifies a location in a communication network. Receiving the search message at the end node containing the data element to a call control means in the communication network so that a connection to the identified location in the communication network can be established. 4. A method according to claim 3, which triggers transmission of the data element. Each pointer includes an identifier associated with a data element stored at the particular end node associated with it, along with a link identifier indicating the next communication link on the route to the end node. 5. The method according to any one of 4 above. The method of claim 1, wherein receipt of the search message at the end node that includes the data element triggers a download of the data element. Means are provided for loading a data element to any end node, and pointer update means is provided for triggering a consequential update of pointers stored in the nodes of the hierarchy in response to the loading of the data element. The method according to any one of claims 1 to 6. A management method for a communication network, comprising a method for accessing a data element according to any one of the preceding claims. The method of claim 8 for use in providing personal numbering services in a communication network. Each pointer along the route identifies a personal number for a particular network user, and the data element located at the end node to which a search message sent along the route has arrived is said in the communication network The method of claim 9, comprising a hardware address for a particular network user.

Images